AI Pulse by Inblix

Topic: gpu-memory

3 articles

Explore our coverage of gpu-memory — 3 curated articles, summaries, and related resources from the Inblix archive.

A KL loss tweak lets a single GPU distill a 120B model without melting — Inblix summary
Research

A KL loss tweak lets a single GPU distill a 120B model without melting

Hugging Face Blog · Aug 10, 2026 · 2 min read

The dirty secret of LLM knowledge distillation isn't the theory—it's the electric bill. Training a smaller model to mim...

PyTorch's Hidden Memory Hog: Why Your 200MB Tensor Needs 3GB During Training — Inblix summary
Research

PyTorch's Hidden Memory Hog: Why Your 200MB Tensor Needs 3GB During Training

Hugging Face Blog · Dec 24, 2024 · 2 min read

That `CUDA out of memory` error isn't just frustrating—it's a lie of omission. It tells you the GPU is full, but not wh...

Quanto FP8 Quantization Cuts SD3 Memory Use by 33% on a Single GPU — Inblix summary
Research

Quanto FP8 Quantization Cuts SD3 Memory Use by 33% on a Single GPU

Hugging Face Blog · Jul 30, 2024 · 2 min read

Running Stable Diffusion 3 in FP16 eats 18.765 GB of GPU memory. That's a problem for anyone without a datacenter card....